Optical networking was entering a period of significant change at OFC 2019. Cloud traffic growth was driving interest in faster data center links, while advances in coherent transmission and modular hardware were creating new options for metro and long-haul networks. At the same time, operators were examining open interfaces and disaggregation as ways to build networks around specific requirements rather than fixed, vertically integrated systems.
These developments are connected, but they address different engineering challenges. Higher data rates can increase capacity per connection; coherent pluggables can simplify some transport deployments; and modular, open systems can change how networks are assembled and managed. Understanding the trade-offs helps technical teams assess which approaches fit their applications and operating models.
400GbE puts cost per bit in focus
At the data center edge, 400GbE was the next major Ethernet speed under development. Optical modules and supporting silicon were being demonstrated, but their cost remained a barrier compared with established 100GbE links. The central commercial question was whether the cost per bit could fall enough to make a migration compelling.
A key enabler was transmission at 100Gbps per wavelength. PAM-4 signaling and digital signal processing (DSP) supported this approach, allowing a 400GbE connection to use four 100Gbps optical lanes. The emerging FR4, DR4, and DR1 formats addressed different link configurations and distances:
- FR4: Four 100Gbps wavelengths over a single fiber pair for links up to 2 km.
- DR4: Four parallel 100Gbps lanes over single-mode fiber for links up to 500 m.
- DR1: A single 100Gbps lane for links up to 500 m, including breakout applications.
One anticipated early use for DR4 was connecting multiple 100GbE servers to a 400GbE switch port. In that scenario, 400GbE supports a denser 100GbE interface before it becomes a native 400GbE connection between network devices. For system architects, this illustrates why an upgrade’s value depends on the overall port and fiber design—not simply the headline link rate.
Coherent optical technology moves toward pluggables
In transport networks, attention centered on lowering the power and packaging requirements of coherent optics while increasing the capacity available on a wavelength. The 400ZR specification targeted 400Gbps links for data center interconnect applications. Related approaches, including ZR+, aimed to extend pluggable coherent technology toward metro and regional distances, while 100ZR addressed shorter-reach network-edge applications.
Pluggable coherent modules can place optical transport capability directly in a switch or router, potentially reducing the need for separate transponder equipment in suitable applications. But reach, cost, module format, and network requirements remain important selection factors. A solution designed for a data center interconnect does not automatically meet the needs of a network with many different link distances.
Vendors were also pursuing coherent systems operating at up to 800Gbps. Raising the baud rate can increase the data carried by each wavelength and may reduce the number of wavelengths and associated equipment required for a given bandwidth. It does not, by itself, increase the total capacity available in a fiber’s optical spectrum. Operators facing fiber-capacity constraints therefore still need to consider spectrum use and fiber availability, not only per-wavelength speed.
Compact modular systems broaden their role
Compact modular platforms began as relatively simple systems for data center interconnect, but were evolving toward more flexible configurations. A single shelf could combine transponders with amplifiers, ROADMs, switching functions, or monitoring and testing capabilities. Sled-based designs also allowed operators to add functions as demand grew instead of purchasing every component up front.
This pay-as-you-grow model can be useful for service providers managing gradual bandwidth growth. Compact systems were also becoming relevant to open line systems and IP-over-DWDM architectures, where colored optics are installed directly in routers or switches rather than in a separate transponder shelf. The fit depends on the optics and functions required: compact platforms can remain useful where a pluggable format or a particular optical function does not fit directly into switching equipment.
Open and disaggregated networks shift complexity
Open networking generally means equipment can interoperate and be managed through standardized interfaces. Disaggregation separates network functions so operators can select hardware for particular applications. The concepts often go together, but they are not identical: a network can be disaggregated without being fully open, and openness does not necessarily require separate hardware components.
Standards-based interfaces, including common data models and control protocols, are intended to make equipment from different suppliers easier to integrate. However, separating functions does not eliminate complexity. It can move that complexity from hardware into software, where operators must coordinate control, provisioning, and management across components. Flexible ROADMs and programmable coherent interfaces make this software and operational work especially important.
Planning for practical network evolution
The developments highlighted at OFC 2019 point to a more varied optical infrastructure rather than a single replacement architecture. Data center teams weighing 400GbE need to consider module economics, port density, and breakout requirements. Transport planners assessing coherent pluggables must match reach and format to each application, while recognizing that higher wavelength rates do not remove fiber-spectrum limits. Compact modular and open systems offer deployment flexibility, but require careful attention to management and integration.
For industrial and network operations teams, the practical lesson is to evaluate each technology as part of the full system: interfaces, power, fiber plant, control software, and growth strategy all matter. Matching those elements to real traffic and operational needs is more valuable than pursuing a higher data rate or a more open architecture in isolation.